Automatic discourse structure detection using shallow textual continuity

  • Authors:
  • Samuel W. K. Chan

  • Affiliations:
  • Department of Decision Sciences, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong

  • Venue:
  • International Journal of Human-Computer Studies
  • Year:
  • 2004

Quantified Score

Hi-index 0.01

Visualization

Abstract

A shallow natural language processing approach to discourse structure detection based on the analysis of textual continuity is described. What distinguishes it from previous research is that it does not work toward on the discovery of the formal subtopic structures. In contrast. attention is focused in uncovering the main factors in textual continuity and simulating a dynamic detection mechanism of cohesive sentence-based fragments. A connectionist filtering algorithln is used to capture the textual continuity as one of the structural backbone of text. As a result, the content conveyed by text with discontinuous topic sequence is. on average, most unlikely to be included in the resultant discourse structure. A prototype and its evaluation with various statistics are included.